Similarity analysis of cardiac signals using temporal features

We present an application of fuzzy clustering to the identification of heartbeat arrhythmias. Many classification algorithms have been developed to distinguish ventricular tachyarrhythmia, a potentially dangerous arrhythmia, from normal sinus rhythm. Few techniques exist to distinguish additional types of arrhythmias due to their overlapping feature sets which prevent successful classification with traditional clustering algorithms. We develop a set of temporal features and a fuzzy clustering algorithm necessary to uniformly and objectively judge the similarity of a patients' arrhythmic heartbeats to their normal sinus rhythm. Our algorithm, called the fuzzy logic similarity analysis algorithm, successfully distinguishes ventricular tachyarrhythmia, as well as ectopic atrial tachycardia, from normal sinus rhythm. We compare our technique to the nearest neighbor clustering algorithm and demonstrate a significant reduction in misclassifications.